Journal of Membrane Science 544 (2017) 68–78

Contents lists available at ScienceDirect

Journal of Membrane Science

journal homepage: www.elsevier.com/locate/memsci

High efficient water/ethanol separation by a mixed matrix membrane MARK

incorporating MOF filler with high water adsorption capacity
⁎ ⁎
Qianqian Li, Quan Liu, Jing Zhao , Yinying Hua, Jiajia Sun, Jingui Duan, Wanqin Jin
State Key Laboratory of Materials-Oriented Chemical Engineering, Jiangsu National Synergetic Innovation Center for Advanced Materials, College of Chemical Engineering,
Nanjing Tech University, 5 Xinmofan Road, Nanjing 210009, PR China

A R T I C L E I N F O A B S T R A C T

Keywords: In order to promote the selective water permeation through porous filler, hydrophilic MOF-801 crystals with
MOF-801 superior water adsorption ability were incorporated into chitosan (CS) matrix to fabricate MOF-801/CS mixed
Mixed matrix membrane matrix membranes (MMMs) for pervaporation dehydration of ethanol. Both the experimental and molecular
Pervaporation simulation results confirm the selective adsorption of water molecules in MOF-801 crystals, while the free vo-
Water adsorption
lume and the lowest energy sorption sites analyses demonstrate the subdued diffusion of ethanol molecules
through MOF-801. As a result, the porous structure in MOF-801 provides additional transport pathways for
water molecules, and meanwhile makes the transport pathways of ethanol molecules more tortuous, thus
achieving simultaneously enhanced flux and separation factor. The optimized membrane with MOF-801 loading
of 4.8 wt% exhibits the total flux of 1937 g/m2 h and separation factor of 2156.

1. Introduction
Polymeric membranes have been extensively investigated and used
in pervaporation dehydration processes of water/alcohol mixtures for
their diversity, low cost, facile preparation, and good film-forming
property and so on. In order to further improve the selectivity and
stability of polymeric membranes, the cross-linking method is usually
adopted, which inevitably reduces the permeation flux due to the de-
creased polymer chain mobility and chain spacing. Alternatively, the
incorporation of rationally designed porous fillers into polymer matrix
to prepare mixed matrix membranes (MMMs) has been identified to be
an effective strategy to address this issue [1–10]. It was proved that the
pervaporation performance can be enhanced by improving the selective
sorption and/or selective diffusion of water molecules via incorporating
appropriate fillers [11–16]. Among various porous materials, the
emerging metal organic frameworks (MOFs) formed through bridging
of metal ions with organic ligands, exhibit high potential as fillers for
molecular separation [17–25]. The existence of organic moieties and
the attendant rich functionalization approaches confer MOFs with
higher interfacial compatibility with polymer matrix compared with
conventional inorganic porous fillers.
To date, diverse MOFs-based MMMs have been employed in per-
vaporation dehydration process [26–29], which mainly focused on
zeolitic imidazolate frameworks (ZIFs) materials such as ZIF-7
(0.29 nm) [9], ZIF-8 (0.34 nm) [30], ZIF-71 (0.48 nm) [31] et al. For
instance, Chung and Amirilargani et al. incorporated ZIF-8 into poly-
benzimidazole or PVA matrix to fabricate MMMs, which achieve sig-
nificantly enhanced permeation flux for alcohol dehydration [30,32]. In
these works, the porous structure provides additional transport path-
ways to increase permeation flux, while the proper pore size smaller
than alcohol molecules confers size-sieving effect. However, ZIFs are
intrinsically hydrophobic, which is disadvantageous for water adsorp-
tion and interfacial compatibility between filler and polymer matrix.
Therefore, it is believed to be effective to promote pervaporation de-
hydration performance via incorporating a new type of MOFs with high
water affinity and especially high water adsorption capacity into
polymer matrix [19,33–38].
In this study, hydrophilic MOF-801 with superior water adsorption
ability was incorporated into chitosan (CS) matrix to fabricate MOF-
801/CS MMMs for pervaporation dehydration of ethanol. MOF-801 is
composed of Zr6O4(OH)4(-CO2) secondary building units, which ex-
hibits stable water adsorption capacity after multiple adsorption/des-
orption cycles [33]. Herein the effects of incorporating MOF-801 on the
morphologies, physical/chemical properties, and PV performance of CS
membrane were systematically investigated. Furthermore the adsorp-
tion behaviors of MOF-801 crystals towards water and ethanol were
intensively studied via experimental and molecular simulation strate-
gies to analyze the impacts of MOF-801 on water/ethanol separation

Abbreviations: E, ethanol; PSI, pervaporation separation index; W, water

⁎
Correspondence to: State Key Laboratory of Materials-Oriented Chemical Engineering, Nanjing Tech University, 5 Xinmofan Road, Nanjing 210009, PR China.
E-mail addresses: zhaojingmem@njtech.edu.cn (J. Zhao), wqjin@njtech.edu.cn (W. Jin).

http://dx.doi.org/10.1016/j.memsci.2017.09.021
Received 1 April 2017; Received in revised form 12 July 2017; Accepted 5 September 2017
Available online 06 September 2017
0376-7388/ © 2017 Elsevier B.V. All rights reserved.
Q. Li et al.
Nomenclature
Symbols
A membrane area (m2)
Ds mass swelling degree (wt%)
E apparent activation energy (kJ/mol)
FAV free volume (dimensionless)
J permeation flux (g/m2 h)
K sorption equilibrium constant (kPa−1)
l membrane thickness (m)
m mass of MOF-801 or polymer (g)
p pressure (kPa)
P/l permeance (GPU)
q equilibrium sorption amount per g of MOF-801 (cm3/g)
Q sorption amount (cm3/g)
So solubility coefficient of permeating molecules
performance.
2. Experimental
2.1. Materials
Fumaric acid was obtained from Alfa Aesar, ZrOCl2·8H2O was pur-
chased from Sigma Aldrich. CS (> 90% N-deacetylation degree) was
obtained from Sinopharm Chemical Reagent Co., Ltd., China. Ethanol
was purchased from Wuxi City Yasheng Chemical Co. Ltd. N,N-
Dimethylformamide (DMF) was obtained from Sinopharm Chemical
Reagent Co. Ltd. Formic acid was purchased from Xilong Chemical Co.,
Ltd. Sulfuric acid and acetic acid (glacial) were obtained from Shanghai
Chemical Reagent Co., Ltd., China. All the above regents are of analy-
tical purity. The flat-sheet polyacrylonitrile (PAN) ultrafiltration
membrane with a molecular weight cut-off of 100,000 was received
from Shanghai Mega Vision Membrane Engineering & Technology Co.
Ltd. (Shanghai, China). Deionized water was used in the all experi-
ments.
2.2. Synthesis of MOF-801
MOF-801 particles were synthesized according to previous report
[33]. Fumaric acid (5.8 g, 50 mmol) and ZrOCl2·8H2O (16 g, 50 mmol)
were dissolved in a solvent mixture of DMF/formic acid (200 mL/
70 mL) in a Teflon-lined stainless steel autoclave and stirred at 303 K
for 30 min to form a homogeneous solution, which was heated at 130 °C
for 6 h. White precipitate was formed and then washed with fresh DMF
and methanol for three times. As-synthesized MOF-801 particles were
immersed in 100 mL of methanol for 3 days, and rinsed with 50 mL of
anhydrous DMF three times per day, during which time the methanol
was also replaced three times per day. Then the solid was dried at
150 °C under vacuum overnight to yield activated MOF-801 sample.
2.3. Preparation of MOF-801/CS MMMs
PAN ultrafiltration membranes were used as supports. A certain
amount of MOF-801 was dispersed in 20 mL of deionized water con-
taining 2 vt% of acetic acid and stirred for about 1 h at room tem-
perature. Then chitosan (0.5 g) was dissolved in the mixture solution
and stirred constantly for 12 h. The solution was filtered to remove the
undissolved residues and kept still for several minutes to remove air
bubbles in the solution. The mixed matrix membranes were prepared
via spin-coating. Before the membrane casting solution was spin-
coated, the PAN ultrafiltration membranes (6 cm × 6 cm) were soaked
in deionized water for 2 days to remove glycerin from the surfaces, and
69
Journal of Membrane Science 544 (2017) 68–78
T feed temperature (K)
t time interval (h)
V cell volume (Å3)
Vo occupied volume (Å3)
W mass of permeate (g)
w mass ratio of MOF-801 to CS (wt%)
WS mass of swollen membrane (g)
WD mass of dry membrane (g)
X mass fractions in feed solution (wt%)
Y mass fractions in permeate solution (wt%)
Greek letters
α separation factor
β selectivity
γ activity coefficient
θ diffraction angle (deg)
then fully dried. The as-prepared MMMs were dried at room tempera-
ture and then immersed into sulfuric acid solution (2 mmol/L) for cross-
linking. After 24 h, the membranes were cleaned with deionized water
and dried at room temperature. A pristine CS membrane was also
prepared using the similar process for comparison. The MOF-801
loading in the membrane was calculated as following formula:
mMOF
wMOF = ×100%
mMOF + mCS (1)
The mixed matrix membranes with the MOF-801 loadings of 0, 2.4,
4.8, 7.0 and 9.1 wt% were named as CS, CM-2.4, CM-4.8, CM-7.0 and
CM-9.1, respectively.
2.4. Characterization
The crystal phases of samples and mixed matrix membranes were
determined by X-ray diffraction (XRD) (Rigaku, Miniflex 600, Japan).
Diffraction patterns were collected at room temperature in the angle
range of 5°≤ 2θ ≤50° with a step width of 0.05° and a scan rate of 0.2 s
step−1. The functional groups of synthesized MOF-801 MOFs were
characterized by Fourier transform infrared (FTIR, AVATAR-FT-IR-360,
Thermo Ncolet, USA) spectra with range of 600–4000 cm−1.
Thermogravimetric analyzer (TGA, STA 209 F1, NETZSCH, Germany)
was used to determine the thermal stabilities of the MOF-801 crystals
and mixed matrix membranes in the temperature range of 30–800 °C
with a heating rate of 10 °C min−1 in a nitrogen atmosphere. The
equilibrium adsorption and desorption isotherms for ethanol and water
vapor with MOF-801 crystals were measured by adsorption instrument
(Belsorp-Max, Bel Inc., Japan). The surface area, micropore volume and
pore size distribution of MOF-801 were calculated from N2 sorption
isotherms using Brunauer-Emmett-Teller (BET), t-plot and NL-DFT
methods (Microtrac BELSORP-mini II, Bel Inc., Japan). The morphology
of composite membrane matrix and topography of membrane surfaces
were observed by field emission scanning electron microscopy (FESEM)
(Hitachi Limited, S-4800, Japan). Energy dispersive spectrometry (EDS)
elemental mapping was taken using an energy dispersion of X-ray
system equipped with a window connected to a FESEMS4800 operating
at 30 kV. The morphology of the crystals was chartered by transmission
electron microscope (TEM) (Tecnai G2 20 S-TWIN). X-ray photoelec-
tron spectroscopy (XPS) (ESCALAB250Xi, ThermoFisher Scientific,
USA) was used to analyze the element composition of the membrane.
2.5. Swelling studies
The homogeneous membranes with different MOF contents were
dried in a vacuum oven. Subsequently, the membranes were immersed
Q. Li et al. Journal of Membrane Science 544 (2017) 68–78

into 90 wt% ethanol aqueous solution at 343 K for 48 h. The membrane The total flux (J, g/m2 h) and separation factor (α) at steady state
surfaces were wiped with tissue paper and then weighed quickly to were calculated by the following equations:
obtain the mass of the swollen membranes. All the experiments were
W
repeated three times. The mass swelling degree (Ds, %) could be cal- J=
At (3)
culated by
Yw /(1−Yw )
Ws − Wd α=
Ds = × 100% Xw /(1−Xw ) (4)
Wd (2)

where Ws and Wd are the mass of the swollen and dry membranes (g). PSI = J (β − 1) (5)

where W (g) represents the mass of permeate over a period of t (h);A

2.6. Pervaporation experiment (m2) is the effective area of flat membrane; l (μm) is the membrane
thickness; Xw and Yw are the mass fractions of water in feed and
The pervaporation experiments were conducted with a homemade permeate side, respectively.
apparatus [41]. The flat membranes were stationed in stainless steel- In addition, the permeance of individual components ((P/l)i, GPU)
made PV cell. The feed solution was maintained at preset concentration (1 GPU = 7.501 × 10−12 m3(STP)/m2 s pa) and selectivity (β) were
and temperature, circulated between the feed tank and the membrane calculated by the following equations:
cell with a flow rate of 12 L/h by a circulation pump. The permeate
Ji Ji
vapor was collected in liquid nitrogen trap. The pressure at permeate (P / l)i = =
pio − pil γi0 x i0 pisat
0 − pil (6)
side was below 300 Pa during collections by a vacuum pump. The
concentration of feed and result were determined by gas chromato-
(P / l)W
graphy (SP-6890, Shandong Lunan, China). The type of packed column β=
(P / l)E (7)
is GDX-102. The detector type is thermal conductivity detector (TCD),
and the detection limit is 10−5 mg/mL. The temperatures of column, where Ji is the permeation flux of component i (g/(m h)), l is the 2

detector and injector are 383 K, 423 K and 423 K, respectively. The PV
performance of a membrane is usually expressed on the basis of the
total flux J and separation factor β. The PSI (Pervaporation separation
index) can be integrated to represent the pervaporation performance,
which is equal to the product of the separation factor α minus one and
the permeate flux J. For each MOF-801 loading, three samples were
fabricated and tested with the average data as the final results.
thickness of membrane (m), pi0, pil are the partial pressures of compo-
nent i in the feed side and permeate side (Pa), and pil can be calculated
approximately as 0 for the high vacuum degree in the permeate side. γi0
is the activity coefficient of component i in the feed liquid, xi0 is the
mole fraction of the component i in the feed liquid, pisat
0 is the saturated
vapor pressure of pure component i (Pa). The permeation flux of water
and ethanol should be transformed into the volumes under standard

Fig. 1. (a) TEM image of MOF-801 crystals; (b) XRD patterns of simulated MOF-801 crystals, as-prepared MOF-801 crystals and acetic acid soaking MOF-801; (c) FT-IR spectra of MOF-
801 crystals; (d) The TGA and DTG curves of MOF-801 crystals.

70
Q. Li et al.
temperature and pressure (STP):1 kg of water vapor at STP = 1.245 m3
(STP), 1 kg of ethanol vapor at STP = 0.487 m3 (STP).
2.7. Simulation model and method
The simulations were carried out in Materials Studio 6.0. First, the
structure file of MOF-801 was downloaded from the CCDC databases.
The fugacity of water and ethanol with different concentration was
obtained by the Wilson calculation [39]. The fixed pressure sorption
and sorption isotherms were performed by the Monte Carlo (MC) [40]
simulation at the above fugacity and temperature. First, the geometric
structure was optimized after energy minimization with 50,000 itera-
tions. Then the sorption simulation was calculated with 106 equilibra-
tion and production steps. The interatomic interaction was described by
COMPASS [41] force field, which is widely used in MOF materials’ si-
mulation. The electrostatic interaction and van der Walls interaction
were predicted by Ewald [42] calculation with the accuracy of
10−5 kcal/mol and atom-based method with 12.5 Å cut-off distance,
respectively.
3. Results and discussion
3.1. MOF-801 crystal characterization
According to TEM image (Fig. 1a), MOF-801 crystals have regular
and uniform particle size about 400 nm. The diffraction peaks from the
XRD pattern of the as-prepared MOF-801 crystals (Fig. 1b) are in good
agreement with the simulated results [38], indicating the high phase
purity. In addition, to identify that the crystal structure was not de-
stroyed in the prepared process of MMMs, MOF-801 crystals were
soaked in acid solution for 12 h. XRD of acid treated MOF-801 crystals
exhibits same diffraction peaks as that of its fresh phase. For the FTIR of
the MOF-801 crystals, the peaks at 1404 cm−1, 1586 cm−1 and
1660 cm−1 are identified as -C˭O-O- bonds, the peaks at 1200 cm−1
and 1100 cm−1 belong to the -C˭O-OH and -C-O- arising from the
functional groups of the fumaric acid in the MOF-801 particles (as
shown in Fig. 1c). Fig. 1d shows the TGA curve of synthesized MOF-801
crystals. A 19.5% weight loss occurred at approximately 110 °C, owing
to the adsorbed water in atmosphere. Another weight loss was at
220 °C, which arised from the evaporation of the DMF solvent from the
particle cages. Furthermore, MOF-801 crystals were thermally stable up
to 400 °C.
N2 physisorption measurement of MOF-801 shows type I isotherm,
characteristic for micro-porosity (Fig. 2a). The specific BET surface
area, Langmuir surface area, micropore volume, and pore size are
986.1 m2/g, 1165.3 m2/g, 0.4751 cm3/g and ~6 Å, respectively. As
Fig. 2. (a) N2 sorption isotherms and pore size distribution curve (insert) of MOF-801; (b) adsorption (solid) and desorption (hollow) isotherms for water and ethanol with MOF-801 at
40 °C.
71
Journal of Membrane Science 544 (2017) 68–78
Table 1
The sorption amounts of simulation and experimental values.
Component Sorption amount (cm3/g)
Experimental values Simulation values
Pure water 330 382.4495
Pure ethanol 20 107.644
shown in Fig. 2b, the water uptake is as high as 350 cm3/g, while
sharply in contrast, the ethanol uptake reaches to 30 cm3/g. The dis-
tinctly different sorption capacities of MOF-801 for water and ethanol
demonstrates its hydrophilic pore surface, reflecting the potential in
solvent dehydration.
To better understand those performance, we used the hard-spherical
particles with given probe radius to formulate the free volume
morphologies of MOF-801-single crystal (MOF-81-SC). MOF-81-SC has
a hollow skeleton and highly interconnected intrinsic microporosity.
The available free volume for water molecules (0.45103) to diffuse is
nearly 1.4 times higher than that for ethanol (0.32711) via fractional
accessible volume (FAV) analysis, especially the windows available for
water connected between two cages. So it can be predicted that water
molecules will permeate faster than ethanol molecules due to the more
free transport space inside MOF-801. In order to verify the reliability of
the simulation, the adsorption performance of the pure component in
MOF-801 was simulated after obtaining the fugacity of pure water and
ethanol at 313.15 K and 1 bar using Wilson method. The results are
shown in Table 1. Both simulated and experimental results showed
preferential adsorption capacity for water molecules. The simulated
water sorption amount presented in Table 1 are in high agreement with
the experimental value, which indicates that the simulation models and
methods used in this work are quite reasonable. The amount of ethanol
adsorption and the experimental value are of some difference, because
the tunnel structure of MOF-801 crystal is rigid while the actual
structure is flexible [43,44]
Free volume can be estimated using the following equations
V − VO
FAVsim . =
V (8)
where V is the cell volume, Vo is the occupied volume.
3.2. Membrane characterization
3.2.1. Chemical and physical structure of mixed matrix membranes
Fig. 3a shows the effect of the MOF-801 loading on the membrane
structure as investigated via XRD characterization. Compared to the
Q. Li et al.
Fig. 3. (a) XRD spectra of different MOF-801/CS MMMs; (b) DTG curves of different MOF-801/CS MMMs; (c) TGA curves of different MOF-801/CS MMMs; (d)XPS spectrum of CM-4.8
membrane.
XRD pattern of pristine CS, the characteristic peaks of MOF-801 at 2θ =
8.5° become stronger with the increasing loading of MOF-801. There-
fore, it can be concluded that the crystal structure of MOF-801 is not
broken during membrane fabrication process, which is consistent with
the result of the stability in acetum solution (Fig. 1b). Fig. 3c shows the
TGA curves of CS control membrane and MMMs. The homogeneous
membranes with different loadings of MOF-801 were dried in a vacuum
oven before measured. The entire thermal degradation process includes
three major weight loss stages. The first stage arises from the eva-
poration of the residual water in membranes (40–200 °C). The pristine
CS membrane shows less weight loss in the first stage compared with
the MMMs, indicating the water adsorption ability of the evacuated
MOF-801 crystals. The second stage is attributed to the deacetylation
and depolymerization of CS (200–300 °C), which is the foremost to
evaluate the thermal stability of membranes. From the TGA and DTG
curves in Fig. 3b and c, both the initial degradation temperatures and
the maximum weight loss temperatures of MMMs are slightly higher
than CS. After incorporating MOF-801 crystals into CS matrix, the hy-
drogen bonds between polymeric and inorganic phases inhibit the
mobility of CS polymer chains, exhibiting a little increased thermal
stability. The third stage belongs to the residual decomposition of CS
(300–800 °C). In summary, all the mixed matrix membranes possess
desirable thermal stability. The XPS spectrum of the CM-4.8 membrane
is shown in Fig. 3d. Zr3d peak appears on the spectrum, confirming the
incorporation of MOF-801 crystals.
3.2.2. Morphology of mixed matrix membranes
Fig. 4a-e shows the surface morphologies of pure CS membrane and
MMMs with different MOF-801 loadings. The surface of pure CS
membrane is smooth and defect-free. After embedding the MOF-801
filler into CS matrix with the loading varying from 2.4 wt% to 9.1 wt%,
72
Journal of Membrane Science 544 (2017) 68–78
the membrane surfaces exhibit granular protuberances without visible
interfacial voids, owing to the good compatibility between MOF-801
and CS matrix. However, with excess MOF-801 crystals (more than
7.0 wt%), agglomeration can be observed in the SEM images of mem-
brane surface.
The cross section of the mixed matrix membrane with MOF-801
loading of 4.8 wt% was characterized by FESEM (Fig. 4f). The thickness
of the membrane is about 1 µm. The elemental maps of the CM-4.8
membrane are shown in Fig. 5. The green-colored, red-colored and
white-colored dots represent nitrogen, zinc and carbon, respectively. Zn
element (red) uniformly disperses in membrane, which indicates the
homogeneous distribution of the MOF-801 nanoparticles.
3.2.3. Swelling study
Swelling is a common phenomenon for hydrophilic membrane in
aqueous solution, which exerts serious influence on membrane struc-
ture and separation performance. Therefore, the mass swelling degrees
of membranes, after being immersed in 90 wt% ethanol aqueous solu-
tion at 343 K for 48 h, were evaluated to characterize the membrane
structure stability (Fig. 6). From the result, the swelling degree of
MMMs slightly fluctuates when the MOF-801 loading is lower than
4.8 wt%, reflecting the negligible influence of incorporating MOF-801.
Although the introduction of MOF-801 crystals in CS matrix may gen-
erate the additional solvent adsorption inside porous filler during
swelling test, the interfacial hydrogen bonds between MOF-801 and CS
restricts the mobility of polymer chains and decreases the adsorption
quantity in polymer matrix, thus offsetting the increased adsorption in
MOF-801. When the MOF-801 loading is higher than 7.0 wt%, the oc-
currence of filler agglomeration generates interfacial microdefects and
lowers the interfacial interaction, leading to increased swelling degree.
Q. Li et al.
Fig. 4. SEM images of different MMMs surface (a-e); (f) cross-section of CM-4.8 membrane.
3.3. Pervaporation experiment
To investigate the effect of incorporating MOF-801 on selective
water-permeation, the pervaporation performance of pristine CS
membrane and MMMs were evaluated with 90 wt% ethanol aqueous
solution at 323 K as shown in Fig. 7. With the increase of MOF-801
loading, the separation factor of MMMs significantly increases at the
beginning and then declines, while the flux exhibits the tendency of
continuous increase (Fig. 7a). When the MOF-801 loading reaches
4.8 wt%, the MMM obtains the optimal performance with the flux of
1113 g/m2 h and separation factor of 2098, which are 11.5% and 160%
higher than those of pristine CS membrane, successfully overcoming the
trade-off effect. After incorporating MOF-801 crystals into CS matrix,
the porous filler with high water adsorption capacity provides addi-
tional transport pathways for water molecules, and meanwhile makes
the transport pathways of ethanol molecules more tortuous. As a result,
73
Journal of Membrane Science 544 (2017) 68–78
water flux increases from 987 g/m2 h to 1110 g/m2 h, while ethanol
flux decreases from 11 g/m2 h to 4 g/m2 h (Fig. 7b). With the further
increase of MOF-801 loading higher than 4.8 wt%, excess MOF-801
crystals may reunite and then lead to microdefects at the interface
between MOF crystals as well as between MOF-801 and CS matrix,
which decrease the diffusion selectivity of membrane towards water,
resulting in the significantly increased ethanol flux and the subse-
quently decreased separation factor. Due to the optimal comprehensive
performance, the MMM with MOF-801 loading of 4.8 wt% was selected
for the following pervaporation experiments.
Pervaporation experiments under different temperatures ranging
from 313 K to 343 K were carried out employing CM-4.8 membrane
with water concentration in feed of 10 wt% (Fig. 8). It can be seen that
increasing the temperature leads to the continuous increase of flux and
separation factor. The high pervaporation performance with the per-
meation flux of 1937 g/m2 h and the separation factor of 2156 could be
Q. Li et al.
Fig. 5. (a) SEM image and (b-d) EDX mapping of (b) N, (c) Zn and (d) C elements from CM-4.8 membrane surface.
Fig. 6. Swelling degrees of different MOF-801/CS MMMs.
obtained when the temperature reaches 343 K. The impacts of opera-
tion temperature on the PV process can be ascribed to the driving force,
the membrane structure and the interactions between permeate mole-
cules and membrane. The continuous increase of water and ethanol
fluxes confirms that the positive impacts on flux arising from the pro-
moted driving force and the loosen membrane structure (i.e. the in-
creased polymer chain mobility) transcend the negative impacts of
weakened interaction between permeate molecules and membrane and
the subsequently reduced molecule adsorption on membrane surface.
The relationship between operation temperature and flux usually
conforms to the Arrhenius law:
−EP
JP = JP 0 exp ( )
RT
Journal of Membrane Science 544 (2017) 68–78
where JP, JP0, EP, R and T represent the permeation flux, pre-ex-
ponential factor, apparent activation energy, gas constant and feed
temperature, respectively. From the slopes of fitted lines in the range of
313–343 K, we can get the activation energies for the permeation of
water (39.19 kJ/mol) and ethanol (17.99 kJ/mol). The much higher
water activation energy indicates the higher temperature sensitivity of
water permeation over ethanol permeation.
The impacts of operation temperature on the PV process can be
ascribed to the driving force, the membrane structure and the interac-
tions between permeate molecules and membrane. In order to achieve
in-depth analysis, the driving force-normalized permeation flux (per-
meance) and selectivity were calculated and listed in Table 2. The
continuously decreased ethanol permeance with temperature confirms
that the positive impact arising from the membrane swelling (the in-
creased polymer chain mobility) is inferior to the negative impacts of
the weakened interaction between water and ethanol (the prohibited
concomitant diffusion of ethanol molecules with water molecules) as
well as the weakened interaction between ethanol and membrane (the
reduced molecule adsorption on membrane surface). On the other
hand, the accelerated water adsorption-desorption process in MOF at a
higher temperature promotes the permeation of water molecules, thus
leading to enhanced water permeance and selectivity.
The effect of feed concentration on the flux and separation factor of
the CM-4.8 membrane at 343 K was also investigated (Fig. 9). It is
shown that when the feed water concentration increases from ap-
proximately 5 wt% to 20 wt%, the permeation flux increases while the
separation factor declines. The increase in the feed water concentration,
normally results in an increasing activity and partial pressure, which
will enhance the driving force for the permeation of water. Besides that,
the aggravated membrane swelling and the consequent larger free vo-
lume cavities arising from higher water concentration also contribute,
which can be confirmed by the increase of water permeance. The
(9)
74
Q. Li et al. Journal of Membrane Science 544 (2017) 68–78

Fig. 7. The pervaporation performance of different MMMs.

ethanol permeance shows more significant increase than water per- Table 2
meance, partly because of the severer influence of membrane swelling Permeance and selectivity of MOF-801/CS membrane under different temperatures and
feed water contents.
on ethanol diffusion, imparting the membrane with higher ethanol
permeability and lower diffusion selectivity at higher water con- Temperature (K) Water content in feed (P/L)W (P/L)E Selectivity
centration. Furthermore, the interaction between water and ethanol (wt%) (GPU) (GPU)
molecules increases with the water concentration increasing from 5 wt
313 10 12,781 12.4 1033
% to 20 wt%, thus promoting the concomitant diffusion of ethanol
323 10 14,694 6.23 2356
molecules with water molecules. As a result, the selectivity and se- 333 10 10,760 4.25 2534
paration factor decline with feed water concentration. 343 10 9964 3.87 2572
343 5 2479 0.34 7282
343 10 3656 0.98 3748
3.4. Sorption isotherms in ethanol/water mixture 343 15 3747 1.31 2869
343 20 4520 2.21 2047
In order to investigate the actual adsorption performance of MOF-
801 in the pervaporation process, the adsorption isotherms of MOF-801
in the water/ethanol mixture solution with water content of 10 wt% at From the limiting slope of the sorption isotherm at zero pressure,
343 K were calculated by the same simulation method. Via fitting the solubility coefficients of water and ethanol can be obtained. Then,
Langmuir type, we obtained the maximal sorption amount (Qi,max) and the solubility selectivity of MOF-801-SC for water can be calculated to
sorption equilibrium constant (Ki,e) of MOF-801 for water and ethanol be nearly 6.0.
molecules (Q H2 O, max = 332 cm3/g, K H2 O, e = 22.051 kPa−1; QEtOH , max = During simulating the realistic sorption process in pervaporation,
144 cm3/g, KEtOH, e = 8.954 kPa−1). The sorption amount of water is the lowest energy sorption sites for water and ethanol are studied and
approximately 2.6 times of ethanol's even though water has a very low presented in Fig. 10. In combination with free volume distribution, we
concentration in feed solution. can clearly find that water molecules will preferentially occupy the
The solubility coefficient of permeating molecules is defined as cages and open windows area, and then exclude the ethanol, while
following [45] ethanol molecules will tend to become clusters in cage's interior (seeing
the magnification area), thus being hindered from jumping out the
q
S0 = lim ( ) windows.
p→0 p (10)
In order to quantitatively explain the higher affinity of MOF-801-SC

Fig. 8. (a) Effect of temperature on the separation performance of the CM-4.8 membrane with water concentration in feed of 10 wt%; (b) Arrhenius plots of permeation flux for separating
water/ethanol mixture by CM-4.8 membrane.

75
Q. Li et al. Journal of Membrane Science 544 (2017) 68–78

Fig. 9. Effect of feed concentration on the separation performance of the CM-4.8 membrane at temperature of 343 K.

to water molecules, we calculate the hydrogen bond in the last frame of

pure water and ethanol sorption in MOF models. Two geometrical
criteria were used to examine hydrogen bonding: (1) the angle of hy-
drogen-donor–acceptor ≤ 150°; (2) the distance between a donor and
an acceptor ≤ 3.5 Å [46]. On average, per water molecule generates
1.43 hydrogen bonds in this MOF skeleton, which is lower than that in
bulk water (3.5)[47] due to the confinement effect occurring in pores of
MOF-801 [48]. While one ethanol molecule forms 0.62 hydrogen
bonds, which indicates that the affinity of MOF-801-SC to ethanol is
much lower than that to water.

3.5. The long-term operation stability

The long-term durability of the pervaporation performance of the

CM-4.8 membrane was tested for 100 h with 90 wt% ethanol aqueous
solution at 323 K. As shown in Fig. 11, the flux fluctuate in a narrow
range while the water concentration in permeate decreases at the first
60 h and then reaches steady state. Similar phenomena were also ob- Fig. 11. The long-time pervaporation performance of the CM-9.1 membrane.

served in previous literatures [49,50], which arises from the plastici-

zation and relaxation process of CS membrane structure under the at-
mosphere of high temperature and solvents and needs to be addressed
in the following research works. After 100 h operation, the stable
membrane performance is still high enough to achieve efficient se-
paration of water/ethanol mixture, indicating the potential application
prospect of MOF-801/CS MMMs.
3.6. Comparison of PV performance with literatures
Chitosan (CS) is one of the most studied materials for pervaporation
dehydration due to its high water affinity for selective sorption, as well
as the structural rigidity and regularity for selective diffusion. Table 2
lists the PV dehydration performance of CS-based MMMs with different
fillers for ethanol/water mixtures. The thinner thickness of the com-
posite membrane is favor of obtaining a higher permeate flux, owing to
the decrease of the transport resistance. In order to comprehensively
compare membrane performance from permeation flux and separation
factor, the PSI values of membranes in literatures and in this work were
calculated. Comparatively, the MOF-801 filled CS membrane prepared
in this work exhibits the highest PSI value and could be a promising
candidate for pervaporation dehydration of water/ethanol mixtures
(Table 3).
4. Conclusions
In summary, MOF-801 nanoparticles with exceptional water

Fig. 10. (a) The density distribution of water and

ethanol and free volume distribution in MOF-801-SC.
(b) Magnification of circular area. Green balls (and
dots) and red ellipse balls (and dots) represent water
and ethanol molecules, respectively. Blue and gray
areas denote the inside and outside surface of cage
voids. (For interpretation of the references to color in
this figure legend, the reader is referred to the web
version of this article.).

76
Q. Li et al. Journal of Membrane Science 544 (2017) 68–78

Table 3
Summary of CS-based MMMs for pervaporation dehydration of water/ethanol mixtures.

Membrane T (oC) Thickness (µm) Water content in feed (wt%) Flux (g/m2 h) Separation factor PSI (× 105 g/m2 h) Reference

CS 25 25 10 602 148 0.89 [9]

CS-OAPS 30 30–40 10 30 373 1.12 [51]
CS-OAPS 60 30–40 10 270 30 0.81 [51]
ZIF-7/CS 25 25 10 322 2812 9.05 [9]
CS-Ti-DHPPA 77 0.25 10 1403 730 10.24 [10]
CS-graphene 30 84 10 10 1093 0.11 [16]
CS-graphene 30 84 10 13 584 0.077 [16]
CS-APTEOS 50 18 15 900 600 5.39 [52]
CS-S-silica 70 30 10 410 1102 4.51 [15]
Prussian blue /CS 25 – 10 614 1472 9.04 [14]
CS-Fe3O4(10) 77 0.6 10 1024 1500 15.36 [13]
ETS-10/CS 50 53 15 550 30 0.17 [53]
CS-MWNT 60 3.27 10 340 573 1.94 [12]
CS-HY zeolite 25 – 10 253 102 0.26 [11]
CS-TiO2 70 – 10 287 207 0.59 [54]
MOF-801/CS 70 1 10 1937 2156 41.76 This work

adsorption capacity were synthesized, and then incorporated into CS
matrix to fabricate MMMs for pervaporation dehydration of ethanol.
Both the experiment and simulation results confirm the preferential
adsorption of MOF-801 for water molecules arising from its higher af-
finity towards water. On the other hand, the free volume and the lowest
energy sorption sites analyses reveal that ethanol molecules tend to
form clusters in the interior of MOF-801 cages, thus the diffusion of
ethanol molecules can be hindered by the water molecules in window
area. Consequently, the introduction of MOF-801 achieves the re-
markably elevated separation factor and flux via simultaneously im-
proving the selective sorption and selective diffusion of water mole-
cules, overcoming the trade-off effect of CS membrane. The membrane
with MOF-801 loading of 4.8 wt% exhibits the highest separation per-
formance with the total flux of 1937 g/m2 h and separation factor of
2156 for water/ethanol mixture with 10 wt% water at 343 K. The ex-
cellent separation performance and easy fabrication process demon-
strate the potential of MOF-801/CS MMMs for pervaporation dehy-
dration of organic solvents.
Acknowledgments
This work was financially supported by the National Natural Science
Foundation of China (21606123, 21490585, 21671102), the Natural
Science Foundation of Jiangsu Province (BK20160980), the Innovative
Research Team Program by the Ministry of Education of China
(IRT_17R54), Six talent peaks project in Jiangsu Province (JY-030), and
Topnotch Academic Programs Project of Jiangsu Higher Education
Institutions (TAPP).
References
[1] J. Shen, G. Liu, K. Huang, Z. Chu, W. Jin, N. Xu, Subnanometer two-dimensional
graphene oxide channels for ultrafast gas sieving, ACS Nano 10 (2016) 3398–3409.
[2] J. Shen, G. Liu, K. Huang, W. Jin, K.-R. Lee, N. Xu, Membranes with fast and se-
lective gas-transport channels of laminar graphene oxide for efficient CO2 capture,
Angew. Chem. Int. Ed. 54 (2015) 578–582.
[3] J. Shen, M. Zhang, G. Liu, K. Guan, W. Jin, Size effects of graphene oxide on mixed
matrix membranes for CO2 separation, AIChE J. 62 (2016) 2843–2852.
[4] S. Yu, Z. Jiang, H. Ding, F. Pan, B. Wang, J. Yang, X. Cao, Elevated pervaporation
performance of polysiloxane membrane using channels and active sites of metal
organic framework CuBTC, J. Membr. Sci. 481 (2015) 73–81.
[5] E.M. Mahdi, J.-C. Tan, Mixed-matrix membranes of zeolitic imidazolate framework
(ZIF-8)/Matrimid nanocomposite: thermo-mechanical stability and viscoelasticity
underpinning membrane separation performance, J. Membr. Sci. 498 (2016)
276–290.
[6] A. Kudasheva, S. Sorribas, B. Zornoza, C. Tellez, J. Coronas, Pervaporation of
water/ethanol mixtures through polyimide based mixed matrix membranes con-
taining ZIF-8, ordered mesoporous silica and ZIF-8-silica core-shell spheres, J.
Chem. Technol. Biotechnol. 90 (2015) 669–677.
[7] I. Erucar, S. Keskin, Computational screening of metal organic frameworks for
mixed matrix membrane applications, J. Membr. Sci. 407 (2012) 221–230.
[8] G. Dudek, M. Gnus, R. Turczyn, A. Strzelewicz, M. Krasowska, Pervaporation with
chitosan membranes containing iron oxide nanoparticles, Sep. Purif. Technol. 133
(2014) 8–15.
[9] C.-H. Kang, Y.-F. Lin, Y.-S. Huang, K.-L. Tung, K.-S. Chang, J.-T. Chen, W.-S. Hung,
K.-R. Lee, J.-Y. Lai, Synthesis of ZIF-7/chitosan mixed-matrix membranes with
improved separation performance of water/ethanol mixtures, J. Membr. Sci. 438
(2013) 105–111.
[10] J. Zhao, F. Wang, F. Pan, M. Zhang, X. Yang, P. Li, Z. Jiang, P. Zhang, X. Cao,
B. Wang, Enhanced pervaporation dehydration performance of ultrathin hybrid
membrane by incorporating bioinspired multifunctional modifier and TiCl4 into
chitosan, J. Membr. Sci. 446 (2013) 395–404.
[11] X. Chen, H. Yang, Z.Y. Gu, Z.Z. Shao, Preparation and characterization of HY
zeolite-filled chitosan membranes for pervaporation separation, J. Appl. Polym. Sci.
79 (2001) 1144–1149.
[12] S. Qiu, L. Wu, G. Shi, L. Zhang, H. Chen, C. Gao, Preparation and pervaporation
property of chitosan membrane with functionalized multiwalled carbon nanotubes,
Ind. Eng. Chem. Res. 49 (2010) 11667–11675.
[13] R. Xing, H. Wu, C. Zhao, H. Gomaa, J. Zhao, F. Pan, Z. Jiang, Fabrication of chitosan
membranes with high flux by magnetic alignment of In situ generated Fe3O4,
Chem. Eng. Technol. 39 (2016) 969–978.
[14] K.C.W. Wu, C.-H. Kang, Y.-F. Lin, K.-L. Tung, Y.-H. Deng, T. Ahamad, S.M. Alshehri,
N. Suzuki, Y. Yamauchi, Towards acid-tolerated ethanol dehydration: chitosan-
based mixed matrix membranes containing cyano-bridged coordination polymer
nanoparticles, J. Nanosci. Nanotechnol. 16 (2016) 4141–4146.
[15] Y.L. Liu, C.Y. Hsu, Y.H. Su, J.Y. Lai, Chitosan-silica complex membranes from sul-
fonic acid functionalized silica nanoparticles for pervaporation dehydration of
ethanol-water solutions, Biomacromolecules 6 (2005) 368–373.
[16] D.P. Suhas, T.M. Aminabhavi, H.M. Jeong, A.V. Raghu, Hydrogen peroxide treated
graphene as an effective nanosheet filler for separation application, RSC Adv. 5
(2015) 100984–100995.
[17] J. Shen, G. Liu, K. Huang, Q. Li, K. Guan, Y. Li, W. Jin, UiO-66-polyether block
amide mixed matrix membranes for CO2 separation, J. Membr. Sci. 513 (2016)
155–165.
[18] Q. Li, L. Cheng, J. Shen, J. Shi, G. Chen, J. Zhao, J. Duan, G. Liu, W. Jin, Improved
ethanol recovery through mixed-matrix membrane with hydrophobic MAF-6 as
filler, Sep. Purif. Technol. (2017).
[19] J. Duan, W. Jin, S. Kitagawa, Water-resistant porous coordination polymers for gas
separation, Coord. Chem. Rev. 332 (2017) 48–74.
[20] S. Kitagawa, R. Kitaura, S. Noro, Functional porous coordination polymers, Angew.
Chem. Int. Ed. 43 (2004) 2334–2375.
[21] M.L. Foo, R. Matsuda, S. Kitagawa, Functional hybrid porous coordination poly-
mers, Chem. Mater. 26 (2014) 310–322.
[22] W.-X. Zhang, P.-Q. Liao, R.-B. Lin, Y.-S. Wei, M.-H. Zeng, X.-M. Chen, Metal cluster-
based functional porous coordination polymers, Coord. Chem. Rev. 293 (2015)
263–278.
[23] S. Pakhira, M. Takayanagi, M. Nagaoka, Diverse rotational flexibility of substituted
dicarboxylate ligands in functional porous coordination polymers, J. Phys. Chem. C
119 (2015) 28789–28799.
[24] R. Haldar, N. Sikdar, T.K. Maji, Interpenetration in coordination polymers: struc-
tural diversities toward porous functional materials, Mater. Today 18 (2015)
97–116.
[25] J.L.C. Rowsell, O.M. Yaghi, Metal-organic frameworks: a new class of porous ma-
terials, Microporous Mesoporous Mater. 73 (2004) 3–14.
[26] Z. Jia, G. Wu, Metal-organic frameworks based mixed matrix membranes for per-
vaporation, Microporous Mesoporous Mater. 235 (2016) 151–159.
[27] W. Li, Y. Zhang, Q. Li, G. Zhang, Metal-organic framework composite membranes:
synthesis and separation applications, Chem. Eng. Sci. 135 (2015) 232–257.
[28] B. Zornoza, C. Tellez, J. Coronas, J. Gascon, F. Kapteijn, Metal organic framework
based mixed matrix membranes: an increasingly important field of research with a
large application potential, Microporous Mesoporous Mater. 166 (2013) 67–78.
[29] Y.K. Ong, G.M. Shi, L. Ngoc Lieu, Y.P. Tang, J. Zuo, S.P. Nunes, T.-S. Chung, Recent

77
Q. Li et al.
membrane development for pervaporation processes, Progress. Polym. Sci. 57
(2016) 1–31.
[30] M. Amirilargani, B. Sadatnia, Poly(vinyl alcohol)/zeolitic imidazolate frameworks
(ZIF-8) mixed matrix membranes for pervaporation dehydration of isopropanol, J.
Membr. Sci. 469 (2014) 1–10.
[31] S. Liu, G. Liu, X. Zhao, W. Jin, Hydrophobic-ZIF-71 filled PEBA mixed matrix
membranes for recovery of biobutanol via pervaporation, J. Membr. Sci. 446 (2013)
181–188.
[32] G.M. Shi, T. Yang, T.S. Chung, Polybenzimidazole (PBI)/zeolitic imidazolate fra-
meworks (ZIF-8) mixed matrix membranes for pervaporation dehydration of alco-
hols, J. Membr. Sci. 415 (2012) 577–586.
[33] H. Furukawa, F. Gandara, Y.-B. Zhang, J. Jiang, W.L. Queen, M.R. Hudson,
O.M. Yaghi, Water adsorption in porous metal-organic frameworks and related
materials, J. Am. Chem. Soc. 136 (2014) 4369–4381.
[34] C. Wang, X. Liu, N.K. Demir, J.P. Chen, K. Li, Applications of water stable metal-
organic frameworks, Chem. Soc. Rev. 45 (2016) 5107–5134.
[35] R. Zhang, S. Ji, N. Wang, L. Wang, G. Zhang, J.-R. Li, Coordination-driven in situ
self-assembly strategy for the preparation of metal-organic framework hybrid
membranes, Angew. Chem. Int. Ed. 53 (2014) 9775–9779.
[36] P. Li, N.A. Vermeulen, X. Gong, C.D. Malliakas, J.F. Stoddart, J.T. Hupp, O.K. Farha,
Design and synthesis of a water-stable anionic uranium-based metal-organic fra-
mework (MOF) with ultra large pores, Angew. Chem. Int. Ed. 55 (2016)
10358–10362.
[37] O.V. Gutov, W. Bury, D.A. Gomez-Gualdron, V. Krungleviciute, D. Fairen-Jimenez,
J.E. Mondloch, A.A. Sarjeant, S.S. Al-Juaid, R.Q. Snurr, J.T. Hupp, T. Yildirim,
O.K. Farha, Water-stable zirconium-based metal-organic framework material with
high-surface area and gas-storage capacities, Chem. Eur. J. 20 (2014)
12389–12393.
[38] Y. Chen, J. Xiao, D. Lv, T. Huang, F. Xu, X. Sun, H. Xi, Q. Xia, Z. Li, Highly efficient
mechanochemical synthesis of an indium based metal-organic framework with
excellent water stability, Chem. Eng. Sci. 158 (2017) 546–551.
[39] G.M. Wilson, Vapor-liquid equilibrium. XI. A new expression for the excess free
energy of mixing, J. Am. Chem. Soc. 86 (1964) 127–130.
[40] D. Frenkel, B. Smit, Understanding Molecular Simulation: From Algorithms to
Applications, Academic press, 2001.
[41] H. Sun, P. Ren, J. Fried, The COMPASS force field: parameterization and validation
for phosphazenes, Comput. Theor. Polym. Sci. 8 (1998) 229–246.
[42] M. Allen, D. Tildesley, Computer Simulation of Liquids, Oxford Science, London,
78
Journal of Membrane Science 544 (2017) 68–78
1990.
[43] N. Lamia, M. Jorge, M.A. Granato, F.A. Almeida Paz, H. Chevreau, A.E. Rodrigues,
Adsorption of propane, propylene and isobutane on a metal-organic framework:
molecular simulation and experiment, Chem. Eng. Sci. 64 (2009) 3246–3259.
[44] H.W.B. Teo, A. Chakraborty, S. Kayal, Evaluation of CH4 and CO2 adsorption on
HKUST-1 and MIL-101(Cr) MOFs employing Monte Carlo simulation and compar-
ison with experimental data, Appl. Therm. Eng. 110 (2017) 891–900.
[45] T.C. Merkel, V.I. Bondar, K. Nagai, B.D. Freeman, I. Pinnau, Gas sorption, diffusion,
and permeation in poly(dimethylsiloxane), J. Polym. Sci. Part B-Polym. Phys. 38
(2000) 415–434.
[46] K.M. Gupta, K. Zhang, J. Jiang, Water desalination through zeolitic imidazolate
framework membranes: significant role of functional groups, Langmuir 31 (2015)
13230–13237.
[47] W.L. Jorgensen, J. Chandrasekhar, J.D. Madura, R.W. Impey, M.L. Klein,
Comparison of simple potential functions for simulating liquid water, J. Chem.
Phys. 79 (1983) 926–935.
[48] P.G.M. Mileo, S. Devautour-Vinot, G. Mouchaham, F. Faucher, N. Guillou,
A. Vimont, C. Serre, G. Maurin, Proton-conducting phenolate-based Zr metal-or-
ganic framework: a joint experimental-modeling investigation, J. Phys. Chem. C
120 (2016) 24503–24510.
[49] C. Gao, M. Zhang, Z. Jiang, J. Liao, X. Xie, T. Huang, J. Zhao, J. Bai, F. Pan,
Preparation of a highly water-selective membrane for dehydration of acetone by
incorporating potassium montmorillonite to construct ionized water channel,
Chem. Eng. Sci. 135 (2015) 461–471.
[50] C. Yeom, J. Jegal, K. Lee, Characterization of relaxation phenomena and permea-
tion behaviors in sodium alginate membrane during pervaporation separation of
ethanol–water mixture, J. Appl. Polym. Sci. 62 (1996) 1561–1576.
[51] D. Xu, L.S. Loo, K. Wang, Pervaporation performance of novel chitosan-POSS hybrid
membranes: effects of POSS and operating conditions, J. Polym. Sci. Part B-Polym.
Phys. 48 (2010) 2185–2192.
[52] J. Ma, M. Zhang, L. Lu, X. Yin, J. Chen, Z. Jiang, Intensifying esterification reaction
between lactic acid and ethanol by pervaporation dehydration using chitosan-TEOS
hybrid membranes, Chem. Eng. J. 155 (2009) 800–809.
[53] C. Casado-Coterillo, F. Andres, C. Tellez, J. Coronas, A. Irabien, Synthesis and
characterization of ETS-10/chitosan nanocomposite membranes for pervaporation,
Sep. Sci. Technol. 49 (2014) 1903–1909.
[54] D. Yang, J. Li, Z. Jiang, L. Lu, X. Chen, Chitosan/TiO2 nanocomposite pervapora-
tion membranes for ethanol dehydration, Chem. Eng. Sci. 64 (2009) 3130–3137.